This paper proposes a motion tracking system to be used for a visual sensing for an autonomous space docking system. The proposed system combines a stereo vision for three-dimensional measuerment and object recognition together with a monocular system for real-time motion tracking. The architecture is carried out using multi-thread parallel processing. The monocular vision thread detects only the feature points of the object using high speed image processing. Then, the three-dimensional features are reconstructed through corresponding each feature point with a point in the previous frame whose three-dimensional coordinate value has already been obtained using stereo measurement. Real-time motion tracking in the three-dimensional space is achieved by repeating this process. A hardware experiment is used to demonstrate the efficiency of the proposed vision system.